9.8
仮説検定の結果により、帰無仮説が棄却されるか棄却されないかが決まります。この決定は、データの分析、適切な検定統計量、適切な信頼水準、臨界値、および P 値に基づいて行われます。しかし、証拠が帰無仮説を棄却できないことを示唆している場合、帰無仮説を「受け入れる」と言うのは正しいのでしょうか?
帰無仮説…
実験では、感染した植物を持つ農場に広く適用可能な殺虫剤を投与します。
この殺虫剤は、施用後、健康な植物の数を増やすことが期待されています。しかし、実験の最後には、健康な植物と感染した植物の割合は同じままでした。
ここでは、殺虫剤には効果がないという帰納仮説が成り立つように思われますが、仮説を受け入れるべきか、それとも否定しないべきか。
この仮説を受け入れると、殺虫剤は効果がなく、植物の健康を改善することはできません。
この決定は、実際には、観察された結果に対する他のもっともらしい説明を見落としています。
この場合、規定されていない量や濃度の殺虫剤を使用しても効果がなかった可能性があります。
殺虫剤では対象とできないものに植物が感染している可能性があります。
帰無仮説を棄却しないということは、期待された効果または観察された効果について十分な証拠がないことを意味します。
今日、科学者が帰無仮説を受け入れていたら、植物ウイルスの発見や多くの絶滅種の再発見は不可能だったでしょう。
View the full transcript and gain access to JoVE Core videos
Q1: Why is 'fail to reject' better than 'accept' when describing null hypothesis test results?
Accepting a null hypothesis implies it is proven true, but hypothesis testing only shows insufficient evidence against it. Failing to reject means the data lacks support for the alternative hypothesis, not that the null is definitively true. This distinction matters because accepting prematurely can halt further investigation and overlook alternative explanations for observed results.
Q2: What are the consequences of accepting rather than failing to reject a null hypothesis?
Accepting a null hypothesis may lead to severe consequences in critical fields like criminal trials, drug testing, and species research. It implies the hypothesis is proven and needs no further study, potentially preventing discovery of viruses, extinct species, or other important findings. Failing to reject leaves room for future evidence to challenge existing conclusions.
Q3: How can alternative explanations affect the interpretation of null hypothesis results?
When a null hypothesis cannot be rejected, multiple plausible explanations may exist beyond the hypothesis being true. In the insecticide example, ineffective results could stem from incorrect dosage, insecticide limitations against specific pathogens, or other factors. Accepting the null overlooks these alternatives, while failing to reject acknowledges the need for further investigation.
Q4: What statistical factors determine whether to reject or fail to reject a null hypothesis?
The decision depends on data analysis, test statistic values, confidence level, critical values and significance level, and P-values. These elements work together to determine if sufficient evidence exists to reject the null hypothesis. When evidence is insufficient, the appropriate conclusion is to fail to reject rather than accept.
Q5: Why can't a null hypothesis be proven true through hypothesis testing?
Hypothesis testing begins by assuming the null hypothesis is true, then evaluates whether data contradicts it. The test can only show insufficient evidence against the null, not prove it true. Absence of evidence against a hypothesis differs fundamentally from evidence proving it true, which is why failing to reject is the correct statistical conclusion.
Q6: How does the insecticide experiment illustrate the difference between accepting and failing to reject?
When plant health remained unchanged after insecticide application, accepting the null would conclude the insecticide is ineffective. However, failing to reject acknowledges other possibilities: incorrect dosage, pathogen resistance, or unsuitable insecticide type. This distinction preserved opportunities for further research that might reveal these alternative explanations.
Q7: What role does newer scientific evidence play in hypothesis testing conclusions?
Newer scientific evidence often challenges existing studies and conclusions. Accepting a hypothesis suggests it is proven and requires no further study, potentially blocking important discoveries. Failing to reject keeps investigations open, allowing future evidence to refine understanding and potentially overturn previous conclusions in fields like virology, paleontology, and medicine.